Forecast error measurement

[Former Member]

Most resources/literature show

Forecast error = Abs (( A - F) / A ) x 100 where A =  Actual, F = Forecast

But some companies follow a different approach as:

Forecast error = Abs ((A-F) / F ) x 100

Under what circumstance you use which formula ? What could be business reasons behind using one over another ?

Thanks,

Susan

[Former Member]

Hi Susan,

I suggest you to use MAPE, well, in fact always dividing by Actual.

You can divide by Forecast but you can get a different approach, because it would mean that you are measuring how the actual value was distant from your forecast it's like saying how wrong was the reality from the idea we had previously.

You should be careful to use the last one, let's suppose Actual Value = 100 and Forecast = 10

MAPE = |100-10|/100 = 90% error

Other(*) = |100-10|/10 = 900% error

Which one of the two measures explain better the situation. Think about it.

[Former Member]

Division by Forecast has the pitfall that planner can adjust the forecast to get the desired value.

While it's not implied that planners do it or will do it, the best method is to divide by actual. In fact, it is not an option to divide by forecast.